Aesthetic Text Logo Synthesis via Content-aware Layout Inferring

Wang, Yizhi; Pu, Guo; Luo, Wenhan; Xiong, Pengfei; Kang, Hongwen; Lian, Zhouhui

doi:10.1109/cvpr52688.2022.00247

Cited by 14 publications

(5 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…According to experiments, the StyleGAN2 model generates lace images with periodicity on one side, but without the cross symmetry. According to Wang et al [23] , We propose a dual discriminator architecture model SStyleGAN.We add a symmetry discriminator. The generator accepts random input to generate images, which are then input into two discriminators for adversarial training.Image discriminator that provides the probability that the image is a real image, and optimizes the generator and image discriminator based on the loss function.…”

Section: Add Symmetry Discriminator Module To Generate Adversarial Modelmentioning

confidence: 99%

SStyleGAN: a StyleGAN model for generating symmetrical lace images

Li,

Zhong,

Chen

2024

International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2023)

View full text Add to dashboard Cite

Lace texture, as a manually designed texture image, needs to possess a series of essential aesthetic characteristics, such as periodicity, symmetry, and blank-leaving in artistic design creation. It requires human designers to spend a lot of time and effort, so it is necessary to apply generative models to generate lace images. In image generation tasks, compared to models such as DCGAN, CycleGAN, and ProGAN, although images generated using StyleGAN2 perform well in terms of resolution, texture details, and periodicity, they still perform poorly in terms of symmetry in lace images. To address the above issues, this article proposes an improved model SStyleGAN (Symmetry StyleGAN) based on StyleGAN2. In terms of discriminators, in order to enhance the attention of the proposed model to image symmetry, we have added a symmetry discriminator, that is, SStyleGAN adopts a dual discriminator structure; In terms of generator, in order to improve the similarity of the feature maps on the left and right sides of the lace image, this paper adds a mean square error loss term based on the loss function of StyleGAN2; In terms of noise input, in order to control the symmetry of the lace image at details such as lace edges, the noise of the StyleGAN2 model is modified to a symmetrical structure, so that the noise input itself has symmetry. In addition to the commonly used FID (Fréchet Insertion Distance) in the generative model, we also used the SSIM (Structural Similarity) metric for the evaluation of the experimental results in this article to detect the symmetry of the generated images. The experimental results show that compared to the lace images generated by the StyleGAN2 model, the lace images generated by the model proposed in this paper not only inherit the advantages of the former, but also have symmetry characteristics.

show abstract

Section: Add Symmetry Discriminator Module To Generate Adversarial Modelmentioning

confidence: 99%

SStyleGAN: a StyleGAN model for generating symmetrical lace images

Li,

Zhong,

Chen

2024

International Conference on Computer Graphics, Artificial Intelligence, and Data Processing (ICCAID 2023)

View full text Add to dashboard Cite

show abstract

“…The wireframe could be described as a differentiable function with regarded to geometric parameters and class probability vectors. In the representation of wireframe, the rectangle's spatial rendering response on (x, y) can be expressed as equation (7) since only the points close to the dotted box's edge (which are in the blue line) are related to the rectangle.…”

Section: Wireframe Rendering Discriminator Architecturementioning

confidence: 99%

“…Figure 11 illustrates the pipeline of Content-aware logo layout GAN with dual discriminator [7] and then further information regarding the layout generating model's pipeline will be provided in the parts that follow. Firstly, we encoded the visual features and character embeddings as a conditioned feature that will be used in the following CGAN.…”

Section: Content-aware Logo Layout Gan With Dual Discriminatormentioning

confidence: 99%

“…In the field of layout design, the newly created, and publicly available datasets in GAN include: the large-scale magazine dataset [5], TextLogo3K [7], and TextSeg [21]. The advertising poster dataset [6] is another large-scale dataset that are can be used for the training model, however, the author has not released the dataset on Github now on the author's commitment that it will be released.…”

Section: Datasetsmentioning

confidence: 99%

“…(3) Text title layout or logo layout design: it is a more specific area compared with the former two areas. It arranges the text and changes the font size according to the semantics of the word to increase its attractiveness and readability [7].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Advances in automatic layout design based on generative adversarial networks

Gao

2023

Sixth International Conference on Computer Information Science and Application Technology (CISAT 2023)

View full text Add to dashboard Cite

Layout is created for elements (images, texts) arrangement which helps show information directly and clearly. Traditional layout design mainly relies on templates or the expertise of experienced designers and usually a good design requires a significantly investment of time and money. Benefitting from the deep learning’s quick development, automatic layout design has progressed based on convolutional neural network especially the Generative Adversarial Networks (GANs). Existing automatic layout design works mainly rely on the framework of Generative Adversarial Networks (GANs), which can further be divided into the graphic layout, components layout and text title layout according to the difference of tasks. Focusing on above three aspects, this paper expounds on all layout design models based on GAN and their application directions and looks forward to future research directions. It can be used as a trade-off and reference for a researcher to choose GAN-based methods when conducting layout.

show abstract